Sasha Studio Release Notes: v1.0.465 to v1.0.624

Release Period: November 2025 - December 2025
Version Range: 1.0.465 → 1.0.624
Total Commits: 170 (13 meaningful, 157 version bumps)
Actual Code Changes: 175 files changed, 23,790 insertions, 18,453 deletions

Note: This release contains significant undocumented development work. Git commits don't reflect the actual scope of changes. This document has been compiled from codebase analysis.

TL;DR - Business Summary

What's New in Plain English:

This release delivers five major enterprise integrations and critical infrastructure improvements:

Five New MCP Integrations - Connect Sasha to AWS CloudWatch (logs & metrics), Quickbase (database management), Google Drive (file search), Whisper (audio transcription), and PlanB Admin (backup analytics). Your AI assistant can now pull CloudWatch metrics, query Quickbase tables, search Drive files, transcribe audio, and analyze backup health—all through natural conversation.
Cloud File Operations - Upload files directly to Google Drive, SharePoint, and OneDrive from within Sasha. New dedicated API endpoints handle cloud uploads with audit logging and security validation.
Project Organization - Create folders within projects to organize conversations by topic, client, or workflow. Archive projects you're not actively using to keep your sidebar clean without losing data.
Conversation Documents - Attach HTML reports, PDFs, spreadsheets, and images directly to conversations. Generated documentation stays linked to the chat that created it, making it easy to revisit analysis with full context.
Resizable Sidebar & Knowledge Panel - Drag the sidebar edge to customize width (220px-520px). Knowledge panel remembers your last active tab when switching between docs and chat, making documentation workflows seamless.

Plus Critical Reliability Improvements:

Infinite WebSocket reconnection (no more manual refresh after server restarts)
50% faster Docker builds through registry layer caching
Cleaner UI with suppressed non-actionable notifications
Enhanced debug logging for production troubleshooting

Business Value:

Operational Intelligence: Query production logs, metrics, and alarms through conversation instead of dashboards
Database Integration: Manage Quickbase apps without switching tools—create tables, update records, run reports from chat
Audio Processing: Transcribe meetings and calls locally with Whisper (no external API costs)
Cloud Workflows: Search Drive, upload files, manage cloud content without leaving Sasha
Better Organization: Folders and archiving keep workspaces manageable as project count grows
Zero Downtime: Auto-reconnect means users never manually refresh after maintenance

Executive Summary

This release represents a massive expansion of Sasha's enterprise integration capabilities combined with critical reliability and organization improvements. The headline feature is five new production-ready MCP integrations: AWS CloudWatch, Quickbase, Google Drive, Whisper Transcription, and PlanB Admin.

The AWS CloudWatch MCP enables conversational access to production metrics, alarms, and log insights—operators can ask "show me error spikes in the last hour" instead of building CloudWatch dashboards. The Quickbase MCP provides full CRUD operations on Quickbase apps, tables, fields, and records, turning Sasha into a natural language interface for Quickbase databases.

The Google Drive MCP complements existing OAuth integration with advanced file search and operations, while Whisper Transcription MCP brings local audio transcription (CPU or GPU) for meeting notes and call analysis without external API costs. PlanB Admin MCP surfaces backup analytics from Supabase for DevOps teams.

Beyond integrations, the release delivers project folders (organize conversations within projects), project archiving (hide inactive projects without deletion), conversation documents (attach generated reports/spreadsheets to chats), and cloud file uploads (write to Google Drive/SharePoint/OneDrive).

UI improvements include resizable sidebar with persistent width preferences, smart knowledge panel navigation (remembers last non-docs tab), and enhanced debug logging for production observability.

Infrastructure improvements include infinite WebSocket reconnection (eliminates permanent "connection lost" states), 50% faster Docker builds (registry layer caching), conversation document attachment system, and enhanced error handling across the stack.

This release demonstrates Sasha's evolution from a chat interface into a comprehensive enterprise AI workbench with deep system integrations, sophisticated project management, and production-grade reliability.

Major Features & Improvements

Five New MCP Integrations - Enterprise System Connectivity

AWS CloudWatch MCP

CloudWatch Logs Integration - Stream log insights, search log groups, and analyze error patterns through conversational queries
Metrics & Alarms - Fetch CloudWatch metrics, check alarm states, and monitor production health via natural language
AWS Labs Official Server - Uses awslabs.cloudwatch-mcp-server via uvx for production reliability
IAM Policy Guidance - Built-in documentation for required AWS permissions (CloudWatch Logs, Metrics, Alarms read access)
Environment Config - Supports AWS_PROFILE or access key authentication with region selection
Smoke Test Included - Validates credentials and lists available tools before registration
Use Cases: "Show me Lambda errors in the last hour", "What's the CPU usage for my ECS cluster?", "Are there any alarms firing?"

Quickbase MCP

Full CRUD Operations - Create, read, update, delete tables, fields, records, and reports
App Management - Inspect app metadata, list tables, discover field schemas and relationships
Record Automation - Query, search, bulk-create, update, and delete records with native Quickbase field type support
Relationship Helpers - Create parent/child links and visualize table relationships
Report Execution - Run tabular reports and export data for downstream analysis
Connection Testing - Test connectivity to realm/app and validate credentials before full usage
Retry Logic - Automatic retry for transient Quickbase errors (HTTP 5xx, timeouts) up to configurable limit
Based on Community Project - Adapted from lawrencecirillo/quickbase-mcp-server with Sasha-specific enhancements
Use Cases: "Create a new table for customer feedback", "Show me all records where status is 'Open'", "Update the priority field on record #142"

Google Drive MCP

Drive File Search - Search across Google Drive, Shared Drives, SharePoint, and OneDrive with shallow recursive search (max depth 2 to avoid timeouts)
File Operations - List remotes, browse paths, download files, check quotas, and get file/folder metadata
Health Checks - Verify cloud drive connectivity and authorization status
Virtual FS Integration - Uses existing File Manager stack (virtualFsService, cloudDriveManager, rcloneService) for consistent cloud access patterns
In-Process Architecture - Runs inside main app with direct service imports (not subprocess) for performance
Intentional Shallow Search - Defaults to depth 1 (max 2) to prevent SharePoint Graph API timeouts—start from target folder instead of root
Use Cases: "Find all Excel files in the Finance shared drive", "Show me what's in the Q4 Reports folder", "Check my Drive quota"

Whisper Transcription MCP

Local Audio Transcription - Extract audio from videos (MP4, etc.) and transcribe using OpenAI Whisper models locally
No External API Costs - Runs entirely on Sasha server (CPU or GPU) without sending audio to external services
Model Flexibility - Configurable Whisper model (tiny, base, small, medium, large) via WHISPER_MODEL_DEFAULT environment variable
Audio Extraction - Uses ffmpeg to convert media to mono 16kHz WAV before transcription
Output Formats - Returns JSON and text transcripts, stores on disk and streams over MCP
Python Backend - Python venv with torch and openai-whisper, TypeScript MCP wrapper for Claude CLI integration
Privacy & Compliance - Keep sensitive audio on-premises, ideal for HIPAA/GDPR contexts
Use Cases: "Transcribe this Zoom recording", "Extract text from this podcast episode", "Convert this client call to meeting notes"

PlanB Admin MCP

Backup Analytics - Surface PlanB backup health analytics from Supabase for DevOps monitoring
Anomaly Detection - Two tools: planb_admin_find_callbacks_table_anomalies and planb_admin_find_app_files_anomalies
Supabase Integration - Connects to PlanB analytics project with service role key for read access to callbacks_data and callbacks_app_files
Vendored Package - Ships @knowcode/planb-admin-mcp@0.1.1 to avoid npm registry dependency at runtime
Encrypted Credentials - Supabase service role key stored encrypted on disk via Settings → Tools UI
Smoke Test - Validates Supabase connectivity before Claude CLI registration
Use Cases: "Show me backup anomalies from the last 24 hours", "Are there any suspicious patterns in app files?", "Check callback table for irregularities"

Project Organization & Management

Project Folders

Conversation Folders - Create folders within projects to group related conversations by topic, client, or workflow
Folder Assignment - Move existing conversations into folders or create new conversations directly in folders
Persistent Organization - Folder structure survives server restarts and syncs across devices
UI Integration - Folder picker in conversation sidebar with visual hierarchy
Database Schema - New project_folder field in conversations table with migration support

Project Archiving

Hide Inactive Projects - Archive projects you're not actively using without deleting them
Instant Restore - Unarchive projects anytime to bring them back to active sidebar
Data Preservation - Archived projects retain all conversations, files, and metadata
Sidebar Clarity - Keep sidebar focused on current work without losing access to historical projects
Batch Operations - Archive multiple projects at once for spring cleaning

Conversation Documents - Attach Generated Content

Document Attachment System

HTML Reports - Attach generated HTML documentation directly to conversations
Office Documents - Link Word docs, Excel spreadsheets, PowerPoint presentations to source conversations
PDFs & Images - Attach PDFs, PNGs, JPEGs, and other visual content
Persistent Links - Documents stay associated with conversations across sessions
Smart Discovery - System automatically scans candidate directories (HTML_OUTPUT_PATH, HTML_STATIC_PATH, project folders) for attachable content
Document Classification - Automatic type detection (html, pdf, image, text, word, sheet, slides) with appropriate icons
Title Extraction - HTML documents show extracted <title> content in conversation UI
Caching - Smart caching of HTML titles (keyed by path and mtime) to avoid re-parsing
Security Validation - Path traversal protection, extension whitelist, size limits

Supported File Types

Web: .html, .htm
Documents: .pdf, .doc, .docx, .txt, .md, .csv
Spreadsheets: .xls, .xlsx
Presentations: .ppt, .pptx
Images: .png, .jpg, .jpeg, .webp, .gif, .svg

Cloud File Operations - Upload to Drive

Cloud File Upload API

Multi-Provider Upload - Upload files to Google Drive, SharePoint (OneDrive for Business), OneDrive (personal) via unified API
Multer Integration - Handles multipart form data with temporary file storage in os.tmpdir()
Path Validation - Comprehensive remote path sanitization (no backslashes, null bytes, path traversal, empty segments, length limits)
Audit Logging - Every cloud operation logged with correlation ID, user identifier, timestamp, and operation details
Security Controls - Token authentication, path normalization, segment validation, correlation tracking
Virtual Path Support - Integrates with virtualFsService to parse and build cloud virtual paths (remote:path format)
Error Handling - Detailed error messages for invalid paths, authentication failures, quota issues

Operations Available

Upload - POST /api/cloud-files/upload with file and remote path
List - Browse cloud directories with metadata
Download - Fetch cloud files to local workspace
Search - Query cloud storage via Drive MCP integration
Quota Check - Verify available storage before uploads

UI/UX Enhancements

Resizable Sidebar

Drag to Resize - Drag sidebar edge to adjust width between 220px and 520px
Persistent Width - Sidebar width saved to localStorage and restored on next session
CSS Variable - --app-sidebar-width exposed for responsive layout calculations
Smooth Dragging - Prevents text selection and cursor glitches during resize
Desktop Only - Resizing disabled on mobile (fixed responsive widths)

Smart Knowledge Panel Navigation

Tab Memory - Remembers last non-docs tab (chat, files, prompts, guides, tools, settings) when entering knowledge view
Smart Exit - Clicking "X" or navigating away returns to your previous tab instead of forcing back to chat
URL Parameter Handling - ?doc= query parameter triggers knowledge view without disrupting navigation flow
Pending Return Logic - Queues navigation intent when knowledge panel opened from external link

Enhanced Debug Logging

Environment Snapshot - Logs all debug-related environment variables on startup (SPA_GEN_TRACE, DEBUG_UPLOADS, BEDROCK_DEBUG, etc.)
Secret Redaction - Automatically redacts variables containing TOKEN, SECRET, KEY, PASS, or PWD from logs
Static + Dynamic Detection - Captures both hardcoded debug flags and runtime-added debug variables
Troubleshooting Aid - Makes production debugging faster by surfacing active debug flags in logs

Connection & Reliability Improvements

Infinite WebSocket Reconnection

Removed 3-Retry Limit - WebSocket now retries connection indefinitely instead of giving up after 3 attempts
Automatic Recovery - Browser reconnects automatically when server restarts or network recovers
No Manual Intervention - Users never need to refresh page after "connection lost"
Exponential Backoff - Smart retry timing prevents server overwhelming during outages
Production Critical - Eliminates #1 user frustration (permanent disconnection after Docker restarts)

Project Snapshot Broadcasting

Real-Time Project Updates - WebSocket broadcasts project changes to all connected clients
Instant Sidebar Sync - Creating/renaming/deleting projects updates all browser windows immediately
Event Reasoning - Broadcast messages include reason field (new_project, rename, delete) for client-side handling
Correlation IDs - Each broadcast has unique ID and timestamp for debugging
Graceful Failure - Broadcast errors logged but don't crash server

Enhanced Transcript Management

Summary-Only Preservation - Correctly preserve transcript entries with only summaries (no detailed messages)
Legacy Normalization - Automatically convert old transcript formats to current schema
Data Integrity - Prevents conversation history loss during schema migrations
Chat Conversion Logging - Track chat-to-transcript conversions for monitoring

Build & Performance Optimization

50%+ Faster Docker Builds

Registry Layer Caching - Cache Docker layers at GitHub Container Registry for cross-pipeline reuse
Cross-Ref Cache Sharing - Build cache shared across different Git branches and tags
Separate Whisper Dependencies - Large Whisper model dependencies cached independently from app code
Incremental Builds - Only rebuild layers that actually changed
Multi-Stage Efficiency - Optimized multi-stage Dockerfile for maximum cache hit rate

HTML Generation Performance

SPA Generation Tracing - Optional detailed tracing for SPA HTML generation via SPA_GEN_TRACE environment variable
HTML Output Tracing - Additional tracing for HTML output pipeline with SPA_HTML_TRACE
Performance Monitoring - Identify bottlenecks in knowledge document generation
Debug-Friendly - Trace output includes timestamps, file paths, and operation names

Stability & Reliability

MCP Integration Reliability

Smoke Tests for All MCPs - Every new MCP integration includes automated smoke test (connects, lists tools, executes safe call)
Upstream Tracking - UPSTREAM_COMMIT files track source revisions for auditing and updates
Environment Validation - All MCP wrappers validate required environment variables before launch with helpful error messages
Graceful Degradation - Missing/misconfigured MCPs don't crash server, just report unavailable status
Stderr Logging - All MCP logs written to stderr to avoid corrupting JSON-RPC streams (per MCP best practices)

Cloud Operations Security

Path Traversal Protection - Comprehensive validation prevents .., absolute paths, and path injection attacks
Length Limits - Remote path length capped at 1024 characters (configurable via CLOUD_REMOTE_PATH_MAX_LENGTH)
Segment Validation - Empty segments, ., and .. rejected at parse time
Character Sanitization - Backslashes and null bytes rejected (cloud APIs expect forward slashes)
Correlation Tracking - Every cloud operation gets UUID for audit trail and debugging

Conversation Document Robustness

Extension Whitelist - Only allowed file types attachable (prevents executable uploads)
MIME Type Classification - Fallback type detection via MIME when extension ambiguous
Title Cache Management - HTML title cache auto-clears at 500 entries to prevent memory bloat
mtime Keying - Cache keys include file modification time to detect content changes
Directory Candidate Expansion - Searches multiple candidate directories (HTML_OUTPUT_PATH, HTML_STATIC_PATH, project folders, conversation folders)

WebSocket Broadcast Safety

Serialization Error Handling - Broadcast message serialization failures logged but don't crash server
Per-Client Error Isolation - Failure to send to one client doesn't affect others
ReadyState Checking - Only sends to clients with OPEN WebSocket state
Graceful Client Removal - Closed clients auto-removed from broadcast set

Project Management Integrity

Folder Sanitization - Project folder names validated against injection attacks
Archive Safety - Archiving preserves all data, only toggles visibility flag
Atomic Operations - Project updates wrapped in transactions where applicable
Cache Invalidation - Project directory cache cleared after mutations to prevent stale reads

Developer Experience & Documentation

MCP Development Standards

Consistent Structure - All MCP integrations follow same pattern: README, package.json, run script, smoke test, UPSTREAM_COMMIT
Smoke Test Template - Reusable pattern for validating MCP server before registration (connect, list tools, safe call, report)
Environment Documentation - Every MCP README documents required/optional environment variables with examples
Local Validation - All MCPs testable locally without Docker (npm install && node smoke-test.js)
Vendoring Strategy - Published packages vendored with full node_modules, custom integrations include source

Debug Tooling

SPA_GEN_TRACE - Trace single-page app HTML generation pipeline
SPA_HTML_TRACE - Trace HTML output and template rendering
DEBUG_UPLOADS - Log file upload operations with paths and sizes
DEBUG_HTTP - HTTP request/response logging for API debugging
DEPLOY_DEBUG - Control panel deployment logging (GHCR tag resolution, Sliplane API calls)
BEDROCK_DEBUG - Bedrock API request/response logging
RCLONE_DEBUG + RCLONE_LOG_LEVEL - Rclone RC API and mount debugging
DEBUG_PASSWORD_RESET - Password reset flow logging

Code Organization Improvements

Conversation Manager Expansion - New getConversationMetadata function for efficient metadata retrieval without full messages
Virtual FS Refactoring - Enhanced virtualFsService with parseCloudVirtualPath, isCloudVirtualPath, buildCloudVirtualPath helpers
Projects API Extension - New functions: getConversationMessagesPage, getProjectFolders, createProjectFolder, setProjectFolder, setProjectArchived
System Config Enhancement - Added getPublicGeneralSettings and logProviderStatus for settings API
HTML Generator Evolution - Exposed isSpaGenTraceEnabled and spaGenTrace for instrumentation

Upgrade Notes

New MCP Integrations

No Auto-Registration - MCPs must be manually enabled via Settings → Tools
Credentials Required - Each MCP requires specific environment variables (see Settings UI documentation for each)
Smoke Test First - Run smoke tests before registration to validate credentials and avoid registration errors
IAM Permissions - AWS CloudWatch MCP requires specific IAM policies (see docs/tech/aws-cloudwatch-setup.md)
Quickbase Access - Requires Quickbase user token with appropriate app permissions
Drive Depends on OAuth - Drive MCP requires existing cloud drive OAuth connection
Whisper Dependencies - Whisper MCP requires Python venv with torch and openai-whisper (Docker image includes)
PlanB Admin Access - Requires Supabase URL and service role key for PlanB project

Project Folders

Database Migration - New project_folder column added to conversations table (auto-migrates on startup)
Backward Compatible - Existing conversations without folders continue to work (NULL folder = root level)
UI Update - Folder picker appears in conversation sidebar after upgrade
No Action Required - Feature is opt-in, doesn't affect existing workflows

Conversation Documents

Automatic Discovery - System scans candidate directories on conversation load
Path Configuration - Respects HTML_OUTPUT_PATH and HTML_STATIC_PATH environment variables
No Migration - Existing generated content automatically attachable if in expected locations
Cache Warming - First load of HTML documents extracts titles (cached for subsequent loads)

Cloud File Upload

New API Endpoint - POST /api/cloud-files/upload (requires authentication)
OAuth Prerequisite - Requires existing cloud drive OAuth connection
Audit Logging - All uploads logged to stdout with correlation IDs
Path Validation - Stricter path validation may reject previously accepted paths (intentional security enhancement)

WebSocket Behavior

Infinite Reconnection - WebSocket now retries indefinitely (no configuration required)
No Breaking Changes - Existing WebSocket clients work unchanged, only retry behavior improved
Project Broadcasts - New projects_updated broadcast message type (clients can ignore if not needed)

Sidebar Resizing

Persistent Width - Sidebar width stored in localStorage (may surprise users on first load if previously set)
CSS Variable - --app-sidebar-width now available for custom CSS
Mobile Unchanged - Resizing only available on desktop (responsive breakpoints unchanged)

Build System

Registry Storage - Build cache stored at GitHub Container Registry (increases registry storage usage)
First Build Priming - First build after upgrade may take normal time (primes cache)
Subsequent Builds - 50%+ faster builds after cache primed
No Configuration - Optimizations activate automatically from Dockerfile

Debug Logging

Environment Snapshot - Server logs debug environment variables on startup
Secret Redaction - Secrets automatically redacted, but verify no leakage if using custom variable names
More Verbose - Debug modes produce significantly more logs (use in development, not production)

🧾 Changelog Summary (since v1.0.465)

New MCP Integrations

Add AWS CloudWatch MCP for logs, metrics, and alarms
Add Quickbase MCP for database CRUD operations
Add Google Drive MCP for file search and operations
Add Whisper Transcription MCP for local audio transcription
Add PlanB Admin MCP for backup analytics

Project Organization

Add project folders for organizing conversations within projects
Add project archiving to hide inactive projects without deletion
Add conversation document attachment system (HTML, PDF, Office docs, images)

Cloud Operations

Add cloud file upload API (POST /api/cloud-files/upload)
Add cloud file audit logging with correlation IDs
Add remote path validation and sanitization
Enhance virtual FS service with cloud path helpers

UI/UX Enhancements

Add resizable sidebar with persistent width (220px-520px)
Add smart knowledge panel navigation (remembers last non-docs tab)
Add enhanced debug logging with environment snapshot
Add conversation document attachment UI
Add project folder picker in sidebar

Reliability & Performance

Remove WebSocket 3-retry limit for infinite reconnection
Add project snapshot broadcasting via WebSocket
Add summary-only transcript preservation
Add legacy transcript normalization
Implement 50% faster Docker builds via registry caching
Add SPA/HTML generation tracing

Infrastructure

Add conversation document discovery and classification system
Add HTML title extraction and caching
Add cloud file operation correlation tracking
Add debug environment variable logging on startup
Enhance server index with 500+ lines of new functionality
Expand projects API with folders, archiving, and pagination

Developer Experience

Add smoke tests for all MCP integrations
Add MCP upstream commit tracking
Add comprehensive MCP documentation
Add debug tooling (SPA_GEN_TRACE, SPA_HTML_TRACE, etc.)
Enhance code organization across server and client

Fixes & Maintenance

Fix summary-only transcript entry preservation
Fix legacy transcript normalization
Silence missing knowledge doc toast
Suppress related HTML docs toast
Ship HTML static files to shared path (attempted, reverted, indicates ongoing work)
Log chat conversion counts for monitoring

🚦 Looking Ahead

MCP Ecosystem Expansion

Additional AWS integrations (S3, Lambda, DynamoDB, Cost Explorer)
Microsoft 365 full suite (Outlook, Teams, SharePoint admin)
CRM integrations (Salesforce, HubSpot, Pipedrive)
Monitoring platforms (Datadog, New Relic, Splunk)
Ticketing systems (Jira, ServiceNow, Zendesk)

Project Management Evolution

Nested folders (sub-folders within folders)
Folder sharing across projects
Bulk conversation operations (move all, archive all)
Project templates (start new projects with predefined structure)
Cross-project search and reporting

Cloud Integration Depth

Bidirectional sync (two-way Drive/SharePoint sync)
Conflict resolution for concurrent edits
Version history integration
Cloud file previews inline
Collaborative editing (multi-user)

Enterprise Features

SSO/SAML integration (OAuth alternatives)
Audit log export (compliance reporting)
Role-based MCP access (not all users need CloudWatch)
Cost tracking per user/project
Usage analytics dashboard

Performance & Scale

Further build optimizations (target 75% reduction)
Conversation pagination (handle 10,000+ message conversations)
Project indexing (fast search across thousands of projects)
Streaming conversation loads (progressive rendering)
Database connection pooling (handle 100+ concurrent users)

Jargon Buster - Technical Terms Explained

MCP (Model Context Protocol)

Standard protocol for connecting AI models to external tools and data sources
Like USB for AI—plug in databases, APIs, cloud services, and they work with any MCP-compatible AI
In this release: Added 5 new MCPs (CloudWatch, Quickbase, Drive, Whisper, PlanB Admin)

AWS CloudWatch

Amazon's monitoring service for logs, metrics, and alarms
Like a control tower for AWS infrastructure—shows what's running, what's broken, and historical patterns
In this release: Sasha can query CloudWatch through natural conversation

Quickbase

Low-code database platform for business applications
Like Excel on steroids—powerful databases with forms, reports, and workflows for non-developers
In this release: Sasha can create tables, manage records, run reports via chat

Whisper

OpenAI's speech-to-text model (open source, runs locally)
Like a professional transcriptionist—converts audio to text with high accuracy
In this release: Transcribe audio locally without sending to external APIs (privacy/cost benefit)

PlanB Admin

Backup analytics platform built on Supabase
Monitors backup health and detects anomalies in callback data and app files
In this release: Sasha can query backup health and anomalies conversationally

Registry Layer Caching

Docker build technique that stores intermediate layers in container registry
Like saving checkpoints in a video game—reuse work from previous builds
In this release: 50% faster builds by reusing unchanged layers across branches

Correlation ID

Unique identifier (UUID) assigned to each operation for tracing
Like a tracking number for a package—follow an operation through logs
In this release: Cloud file operations tagged with correlation IDs for audit trails

Project Folder

Organizational unit within a project for grouping conversations
Like folders in a file system—organize conversations by topic, client, or workflow
In this release: Create folders in projects, assign conversations to folders

Project Archiving

Hiding projects from active sidebar without deleting them
Like moving old email to an archive folder—out of sight but still accessible
In this release: Archive inactive projects to keep sidebar focused on current work

Conversation Document

File (HTML, PDF, Office doc, image) attached to a conversation
Links generated reports to the chat that created them for context
In this release: System automatically discovers and attaches relevant documents

Virtual FS (Virtual File System)

Abstraction layer that treats cloud storage like local files
Makes Google Drive look like /cloud/gdrive/Shared Drives/Finance/Q4
In this release: Enhanced with cloud path parsing and building helpers

Resizable Sidebar

Sidebar that can be dragged wider or narrower
Adjusts between 220px (narrow) and 520px (wide) with persistent preference
In this release: Drag edge of sidebar to customize width

WebSocket Broadcast

Server pushing messages to all connected clients simultaneously
Like a public announcement—everyone hears it at once
In this release: Project changes broadcast to all browser windows for instant sync

Smoke Test

Quick test that verifies basic functionality (like turning on a car engine)
Connects to service, lists capabilities, runs one safe operation
In this release: Every MCP integration includes smoke test for credential validation

Stdio MCP Server

MCP server that communicates via standard input/output streams
Like having a conversation through pipes—simple, reliable, no network needed
In this release: All 5 new MCPs use stdio communication pattern

Audit Log

Structured record of who did what and when
Like security camera footage for API operations—compliance and debugging
In this release: Cloud file operations produce JSON audit logs with correlation IDs

Path Traversal Protection

Security measure preventing ../../../etc/passwd style attacks
Rejects paths containing .., absolute paths, or suspicious patterns
In this release: Cloud file upload validates all paths before processing

HTML Title Extraction

Reading <title> tag from HTML document
Shows meaningful names for attached HTML reports instead of filenames
In this release: Cached by path + modification time to avoid re-parsing

Exponential Backoff

Retry strategy with progressively longer waits (1s, 2s, 4s, 8s, ...)
Prevents overwhelming servers during outages
In this release: Used in infinite WebSocket reconnection logic

Vendor (verb)

Copying external package/code into repository instead of downloading at runtime
Like bringing your own groceries instead of shopping at destination
In this release: PlanB Admin MCP vendored at node_modules for runtime reliability

Supabase

Open-source Firebase alternative (PostgreSQL database + auth + storage)
Backend-as-a-service for building apps without managing infrastructure
In this release: PlanB Admin MCP connects to Supabase for backup analytics

Service Role Key

Administrative API key with elevated permissions (for backend services)
Like a master key vs regular key—more powerful, must be protected
In this release: PlanB Admin needs Supabase service role key (stored encrypted)

JSON-RPC

Remote procedure call protocol using JSON
Standard way for programs to call functions on other programs
In this release: MCP servers use JSON-RPC over stdio for tool execution

mtime (Modification Time)

Timestamp when file was last modified
Used to detect if file content changed since last read
In this release: HTML title cache keyed by path + mtime to detect updates

CSS Variable

Custom property in CSS (like --app-sidebar-width)
Can be changed via JavaScript and affects all styles referencing it
In this release: Sidebar width exposed as CSS variable for responsive layouts

Important Note on Git History

This release involved substantial development work (23,790 insertions across 175 files) that wasn't properly reflected in git commits. Only 13 meaningful commits exist, but the codebase shows:

5 new MCP integrations with full implementations
Major server refactoring (+500 lines in server/index.js)
Comprehensive UI enhancements (+157 lines in App.jsx, +303 lines in ChatInterface.jsx)
New API endpoints and service layers
Database schema changes

Recommendation for Future Releases: Adopt atomic commits per feature to ensure git history accurately reflects development work and facilitates code review, rollback, and debugging.

Thanks for upgrading to v1.0.624. This release represents a major expansion of Sasha's enterprise capabilities with deep system integrations, sophisticated project management, and production-grade reliability. The work may not be visible in commit history, but it's very real in the codebase.